US 7133365 B2
A system and method for controlling routing of data. It is determined whether at least one data flow is active, the at least one data flow having an ingress portion of a current ingress path to a first point from a second point and an egress portion of a current egress path to the second point from the first point. Another data flow is routed via a next ingress portion of a next path to the first point from the second point.
1. A method for controlling data flow in a multi-homed network, the multi-homed network including a source connected to a destination through a plurality of data network connections, the method comprising:
identifying cost factors for a first data flow, the cost factors including one or more of carrier data, contract data, list usage data, or billing estimate data;
passively monitoring path performance for a path associated with the first data flow to obtain path performance information;
identifying a first ingress path for the first data flow from the destination to the source based at least in part on the cost factors and path performance information;
identifying a first data network connection within the plurality of data network connections that includes a portion of the first ingress path; and
routing the first data flow through the first data network connection so that the first data flow is routed along the first ingress path.
2. The method of
3. The method of
4. The method of
5. The method of
6. The method of
7. The method of
8. The method of
9. The method of
10. The method of
11. A method for routing data in a network comprising a plurality of data network connections between a first point and a second point comprising:
passively monitoring performance characteristics of a first data flow along a first path;
determining an activity status of the first data flow based on the path performance characteristics;
if the activity status of the first data flow is active, then preserving a first ingress portion of the first path;
if the activity status of the first data flow is inactive, then determining that the first ingress portion is not available to route data.
12. The method of
forming an ingress association for the first data flow and the first ingress portion; and
routing the first data flow via the first ingress portion in accordance with the first ingress association.
13. The method of
identifying the first data flow with at least one remote destination address related to the second point;
identifying the first ingress portion with a first local network address related to a local source address;
associating the at least one remote destination address with the first local network address; and
storing data representing the association between the at least one remote destination address with the first local network address in a data structure within a database.
14. The method of
preserving a first egress portion for the first data flow, if the activity status of the first data flow is active; and
determining that the first egress portion is not available to route data, if the activity status of the first data flow is inactive.
15. The method of
forming an egress association for the first data flow and the first egress portion; and
routing data via the first egress portion in accordance with the first egress association.
16. The method of
identifying the first data flow with at least one remote destination address related to the second point;
identifying the first egress portion with a first interface;
associating the at least one destination address with the first interface; and
storing data representing the association between the at least one remote destination address in a data structure within a database.
17. A system for controlling routing of data through one or more data networks, the system comprising:
a flow inspector designed to receive local network data traffic from or to a local source address, the flow inspector configured to determine that at least one data flow of the local network traffic is active by passive flow analysis;
an information base coupled to the flow inspector to receive data representing one or more flows that are active, the information base configured to provide a first ingress portion of a first ingress path for a first data flow from a second point to the first point; and
a data director coupled to the information base to receive the local network data traffic and data representing a second ingress portion of a second ingress path, the data director configured to route data via the first ingress portion for active flows and to route data via the second ingress portion for next data flows.
18. The system of
This application claims priority from U.S. Provisional Application No. 60/350,186 entitled “SYSTEM AND METHOD TO ASSURE NETWORK SERVICE LEVELS AND BANDWIDTH MANAGEMENT WITH INTELLIGENT ROUTING,” filed on Nov. 2, 2001, and incorporated herein by reference for all purposes. In addition, U.S. patent application entitled “SYSTEM AND METHOD TO ASSURE NETWORK SERVICE LEVELS WITH iNTELLIGENT ROUTING,” having U.S. patent application Ser. No. 09/833,219, filed on Apr. 10, 2001, and U.S. patent application entitled “SYSTEM AND METHOD TO PROVIDE ROUTING CONTROL OF INFORMATION OVER DATA NETWORKS,” having U.S. patent application Ser. No. 10/013,809 filed on Dec. 7, 2001 are incorporated herein by reference for all purposes.
The present invention relates generally to routing of data over networked communication systems, and more specifically to controlled routing of data across networks, by modifying an ingress path into or an egress path from a local source network, or both.
One traditional technique used to control routing of information over data networks, such as the Internet, uses two or more data network connections to form a multi-homed local source network. These multiple connections increase the bandwidth or throughput of the amount of data capable of traversing the networks from an originating local source network or source address to a remote destination. With increased bandwidth, performance and reliability of Internet traffic is improved. The multi-homed connections to the Internet generally are across several different network service providers.
Conventional multi-homing schemes typically use Border Gateway Protocol (BGP) to direct traffic across one or more network service providers' links. Using traditional multi-homing with BGP effectuates route control by making changes to a routing table such that only outbound traffic from the source network to a remote destination is modifiable.
One drawback of using conventional multi-homed route control is that inbound traffic from the remote destination to the source network is not available for route control using conventional BGP multi-homed techniques. For example, route control is realized by first determining the next hop of a remote destination address and then by changing that next hop address. Although the path from which data traffic leaves the multi-homed destination can be controlled, there is little control of the inbound path into which traffic enters a multi-homed local source network.
Additionally, multi-homing with BGP is a complicated endeavor. There is a need for a wide variety of complicated routing policies to implement BGP multi-homing and thus require specific BGP expertise. Many entities using local source networks that would benefit with multi-homing, however, do not have the necessary expertise to effectively utilize BGP multi-homing. Furthermore, multi-homing deployed in a only small number of destination sites on a data network, such as the Internet, requires injecting long prefix address blocks into one or more network service providers. This means that global routing tables needed for route control are becoming cumbersome because they are growing in size at an alarming rate on the Internet. This has the effect of hindering network performance by bogging down backbone routers as well as other network equipment.
To address the problems of traditional BGP multi-homing, a conventional approach uses network address translation (“NAT”) along with a multi-homing scheme. In this conventional method, two or more blocks of address space from an upstream service provider (e.g., upstream from an originating local source network) are used in a NAT-based routing policy. The NAT policy directs traffic into a particular service provider to avoid problematic providers.
One drawback to this approach is that network path diversity is not guaranteed. Since an upstream service provider advertises only a particular small address block used, the data traffic is certain to enter the multi-homed site via the provider of the original NAT source address space. Another drawback means that multi-homing with NAT must be fairly static. That is, without knowing the history of routing information with NAT, NAT-based multi-homing cannot change a specific NAT source address in the middle of a TCP flow because it would disrupt that flow and cause service outages. The requirement that multi-homing be static with NAT prevents NAT from being used in any sort of intelligent routing or route control system today. Since intelligent routing or route control systems are limited to BGP and hence outgoing route control only, they are only applicable to content distribution (e.g., asymmetric data flows between a source and a destination) where the majority of the traffic is outbound from a multi-homed destination or source.
Therefore, conventional multi-homing with NAT is not suitable to control routing of symmetric data traffic between a local source and remote destination, such as VPN, voice, video, or business applications such as customer relationship management programs (CRM), supply chain management software, or any application where the traffic flows in a bi-directional manner. Thus, traditional route control systems are not capable of influencing the inbound path of the network alone, or both the inbound and outbound paths.
There is a need in the art for a system and a method to overcome the above-described shortcomings to effectively and efficiently control routing of data over multiple networks. Accordingly, there is a need to provide data network users, such as Internet users, with control over routing of data by optimizing egress data paths from a local host associated with a local source address, ingress date paths into the local host, or both.
In one embodiment according to the present invention, a method facilitates controlling routing of data. The method comprises determining at least one data flow is active, the at least one data flow having an ingress portion of a current ingress path to a first point from a second point and an egress portion of a current egress path to the second point from the first point; and routing another data flow via a next ingress portion of a next path to the first point from the second point.
In another method according to the present invention, a system facilitates controlling routing of data through one or more data networks. The system comprises a flow inspector designed to receive local network data traffic from or to a local source address, the flow inspector configured to determine that at least one data flow of the local network traffic is active; an information base coupled to the flow inspector to receive data representing one or more flows that are active, the information base configured to provide a current ingress portion of a current ingress path for the at least one active flow from a second point to first point; and a data director coupled to the information base to receive the local network data traffic and data representing a next ingress portion of a next ingress path, the data director configured to route data via the current ingress portion for active flows and to route data via the next ingress portion for next data flows.
Detailed descriptions of various embodiments are provided herein. It is to be understood, however, that the present invention may be embodied in various forms. Therefore, specific details disclosed herein are not to be interpreted as limiting, but rather as a basis for the claims and as a representative basis for teaching one skilled in the art to employ the present invention in virtually any appropriately detailed system, structure, method, process or manner.
As with the external physical configuration shown in
The exemplary use of the Internet for distribution or communication of information is not strictly necessary to practice the present invention but rather is merely used to illustrate a specific embodiment. Further, the use of server computers and the designation of server and client machines are not crucial to an implementation of the present invention. USER1 Computer can be connected directly to the Internet. Server1's connection to the Internet is typically by a relatively high bandwidth transmission medium such as a T1 line, a T3 line, Metro Area Ethernet, or the like, although it might be connected in a similar fashion as with USER1. Similarly, other computers 84 are shown utilizing a local network (e.g., Local Area Network, or LAN) at a different location from USER1 Computer. The computers at 84 are coupled via Server2 to the Internet. Although computers 84 are shown to include only a single server (e.g., Server2), two or more servers can be connected to the local network associated with computers 84. The USER3 and Server3 configuration represent yet a third network of computing devices.
An exemplary data network includes one or more source networks 94. A source network 94 typically is a local network including one or more servers owned and operated by application service providers, managed service providers, content delivery networks, web hosting companies, individual enterprises, corporations, entities and the like. Such service providers typically communicate information to users that are further removed from the multi-homed NSPS 92, such as NSP 1, NSP 2, NSP 3, . . . and NSPn. In one example, NSPS 92 are coupled to a source network or source point as to be considered a first set of data networks. These NSPs, or first set of data networks, are in turn coupled to a second set of networks, wherein the second set is connected to multiple other networks, thus establishing one or more paths from a source to a destination. A path as described herein can be a route from a source to a destination that is divided into segments, each segment residing wholly within a provider.
The multiple connections between router 98 and multiple network service providers 92 provide an operator of source network 94 a means with which to direct data traffic according to the best performing network service provider. Switch 96 operates to transfer bi-directional data 99, such as IP data, bi-directionally from source network 94 to router 98. Although a single router and switch are shown, one having ordinary skill in the art will appreciate that either additional routers and switches or other suitable devices can be substituted according to another embodiment of the present invention. Moreover, switch 96 need not be used to practice the subject invention. In a specific embodiment, router 98 includes one or more routers running an exemplary protocol, such as BGP (e.g., BGP4, such as Cisco™ or Juniper implementations™), for example, and preferably has route visibility across multiple network service providers.
In an embodiment of flow control system 90, system 90 operates to measure end-to-end (i.e., source-to-destination and destination-to-source) data traffic 95 in terms of flow characteristics, such as performance, cost, bandwidth, and the like. Flow control system 90 also generates statistics associated with data paths across multiple network service providers in real-time, or near-real-time. Such statistics are communicated to source network 94 for providing network engineering personnel, for example, with report information 91 such that immediate reports are created to provide information related to route-change activity, traffic performance as delivered to selected destinations and transit provider usage (i.e., bandwidth), cost, and the like.
In one embodiment according to the present invention, a local computing device uses report information 91 from system 90 to generate visual and graphical representations on, for example, a user-friendly interface (UI) where the representations are indicative of data traffic along one or more paths (e.g., paths between a source and a destination). Network personnel, or any entity responsible for flow control, with access to source network 94 then can provide control information 93 to flow control system 90 to modify system operation by, for example, changing data traffic flow from a under-performing current, or default, path to a better performing path. Intervention by network personnel, however, is not necessary for flow control system 90 to operate in accordance with the present invention.
Flow control system 90 further functions to compare specific data traffic flows (i.e., both uni- and bi-directional traffic flows outbound from and inbound into the data network) to determine whether a particular traffic flow meets one or more rules of an associated flow policy. A flow policy, as referred to herein, includes a set of one or more rules that is associated with a particular data traffic flow related to particular system user (e.g., as denoted by an IP address prefix).
A rule, or criterion, is a minimum level, a maximum level or a range of values that defines acceptable routing behavior associated with a traffic flow characteristic. For example, a rule can set: the maximum acceptable cost, with or without regard to network service provider cost; the maximum load or bandwidth usage associated with traffic flows through specific providers; a range of acceptable (or non-acceptable) service providers; the maximum acceptable latency or loss over one or more paths across multiple network service providers; acceptable ranges of performance for each NSP, such as maximum burst limits, minimum performance commitments and range of costs (i.e., cost structures with regards to time of day, type of traffic, etc.); and any other data flow characteristic that can influence the measurement or the control of data traffic.
Flow control system 90 further operates to detect when one or more rules, or flow policies, are violated and then takes remedial action. That is, flow control system 90 enforces policies associated with data traffic flow by correcting detrimental deviations in performance (i.e., service level assurance), costs or bandwidth (i.e., load in terms of percent capacity available per path). Flow control system 90 makes such corrections based on real- or near-real time traffic analysis, local path diversity (i.e., modifying one or more egress paths from a data network), and visibility into downstream available paths. For example, for a destination related to a specific traffic flow, flow control system 90 directs, or re-directs, traffic to one or more alternative paths to resolve a particular flow's deviation in terms of flow characteristics, from its flow policy.
Flow control system 200 includes controller 205, passive calibrator 203, active calibrator 208, configuration element 211, and usage collector 214, each of which can be realized in hardware, software, or a combination thereof. For example, controller 205, passive calibrator 203, active calibrator 208, configuration element 211, and usage collector 214 are software modules designed to perform specific processes, as described herein, in accordance with the present invention. Such modules can reside in one or more computing devices, such as the computing devices shown in
Exemplary passive calibrator 203, active calibrator 208 and usage collector 214 are coupled to controller 205 to, in part, provide flow characteristics of data traffic. Controller 205 receives monitored flow characteristics as well as flow policies to be enforced. Controller 205 is configured to determine if a flow policy is violated, and upon detection of such a violation, to then select a remedial action to resolve the violation. Configuration element 211 is coupled to controller 205 used to receive information to initiate remedial actions and is configured to communicate such actions to data director 220. Thereafter, data director 220 implements the corrective action to resolve the pending violation, for example, by changing the traffic flow from the current path to a better performing path.
Additionally, flow control system 200 includes traffic repository 221 and flow policy repository 218. Exemplary traffic repository 221 and flow policy repository 218 are databases, such as storage devices configured to store a large number of records in one or more data structures. Traffic repository 221 is designed to store and to communicate information related to traffic and route characteristics, and flow policy repository 218 is designed to store and to communicate policy information or rules to govern the performance and cost of each of the data traffic flows. One having ordinary skill in the art of database management should appreciate that many database techniques may be employed to effectuate the repositories of the present invention.
In operation, flow control system 200 of
Passive calibrator 203 monitors the data traffic of data flow 201 and communicates information 204 related to the traffic and traffic performance to controller 205. Controller 205 is configured to receive policy data 206 representing one or more policies that correspond to a particular traffic flow, such as a particular data flow. Moreover, the particular data flow can be associated with a certain user identified by a destination prefix, for example. From policy data 206, controller 205 determines the levels of performance, cost, or utilization that the particular traffic is to meet. For example, controller 205 determines whether a particular traffic flow of data flow 201 is meeting defined performance levels (i.e., service levels) as defined by one or more requirements or criteria, such as inbound and outbound network latency, packet loss, and network jitter.
Active calibrator 208 functions to send and to receive one or more active probes 207, of varying types, into and from the data networks. These probes are designed to measure network performance, including the path taken across one or more available providers (i.e., to determine if a provider is a transit AS rather than peer AS), next hop-in-use, and other network parameters. To activate active calibrator 208, controller 205 sends an active probe request 209 to active calibrator 208. Such a request is required if controller 205 determines that additional information regarding alternative paths or network system characteristics are necessary to better enforce policies in reactive flow control systems or Alternatively, such request may be required to prevent the occurrence of such policy violations in optimized flow control systems.
Usage collector 214 is configured to receive NSP data 217 representing one or more network provider configurations. Generally, such configurations include the number of paths (“pipes”) associated with each provider and the size thereof. Additionally, NSP data 217 can relate to a provider's cost or billing structure and can also include each provider's associated set or sub-set of addresses, each provider's billing methods (i.e., byte/min, etc.), etc. Moreover, usage collector 214 is configured to collect usage information 213 from the network elements, such as switches, border routers, provider gear, and other devices used to transport data over data networks. Usage collector 214 is configured to provide controller 205 with provider utilization and billing information 215, which represents aggregated data based upon NSP data 217 and usage information 213. Utilization and billing information 215 includes data that represents cost, billing, utilization, etc., for each network service provider of interest.
One having ordinary skill in the art should appreciate that NSP data 217 can be provided to usage collector 214 in a variety of ways. For example, the data can be provided by the data paths used by the data flows or can be provided by an entity having authority to do so, such a network engineer entering the data into a computing device in source network 94 of
Moreover, usage collector 214 is configured to monitor usage characteristics defining a NSP's data traffic capacity, costs, etc. Usage information 213 provided to usage collector 214 includes usage characteristics from network elements, such as switches, border routers, routers, provider gear, and other devices used to transport data over data networks. Usage refers to the data (i.e., raw data such as X Mb samples at time(0)) that represents instantaneous or near instantaneous measurement of characteristics (i.e., usage characteristics) that define, for example, the load and available capacity of each network service provider. “As used herein, Utilization” the usage rate (i.e. the Change in usage over time). For example, suppose the usage collector monitoring NSP1 measures its utilization, or capacity over time, as X Mb at time (0) and Y Mb at time(1). This raw data, or usage, is used to calculate utilization, or usage rate for NSP1 (e.g., Y–X/time(1)-time(0)). Bandwidth is the total capacity of each path or segment of path available for traffic flow. In one embodiment, the usage can be measured in any segment in any path at any number of hops or networks from a first point. Load is typically defines the amount of capacity of a particular path that is used to carry data traffic and can be expressed as load/bandwidth.
Usage collector 214 is designed to generate utilization and billing information 215 based upon usage information 1213 and NSP data 217. Since each of the providers has different cost and billing structures, as well as methods of determining usage costs, usage collector 214 operates to aggregate usage information 213 accordingly to provide controller 205 with utilization and billing information 215.
Usage collector 214 then provides the utilization billing information 215 to controller 205 for each NSP of interest. One having ordinary skill in the art should appreciate that the usage collector can provide additional information based upon the provider usage information, to the controller, as needed to better effectuate route control.
Controller 205 collects information (i.e., aggregated performance and usage characteristics) from each of passive calibrator 203, active calibrator 208, usage collector 214, and optionally traffic repository 221. Based upon the information collected, controller 205 determines a course of action that best alleviates the policy violations in respect to the information represented by policy data 206 that is conveyed to controller 205. Once the coarse of action is determined, controller 205 initiates and sends a network routing change request 212 to configuration element 211. In a specific embodiment, controller 205 also provides data representing one or more alternate data paths that can be used to resolve the policy violation.
Configuration element 211 is designed to communicate routing changes in the network to data director 220. Once configuration element 211 sends one or more routing changes, data director 220 then moves data flow 201 from a current path to another path (e.g., from NSP1 to NSPn or a first path of NSP1 to a second path of NSP1). Data director 220 thus operates to distribute traffic to these destinations across multiple NSP links based on, for example, the cost and performance measured across each link.
In operation, configuration element 211 communicates one or more routing changes 210 with data director 220, for example, by using a routing protocol such as BGP. Configuration element 211 functions to dynamically control routing behavior by modifying the source address of the traffic passing through data director 220. The source address is modified in a way that improves application performance as well as cost requirements.
The following discussion includes more description of each of the elements of exemplary control system 200. Referring back to active calibrator 208, active calibrator 208 provides active mechanisms within system 200 for determining the nature of downstream or upstream paths. This information is typically not available in any conventional protocol used on data networks such as the Internet, and must be collected beyond the information generally available. As shown in
In one embodiment, active calibration probes are communicated to available network or Internet paths via probe path 207. The returning active calibration probes enter via probe path 207 into active calibrator 208. Active calibrator then forwards probe information 209 to controller 205, which contains performance information including alternate available paths. Controller 205 then determines how best to enforce the specifics of the policy associated with the subject traffic flow. Exemplary active calibrator 208 employs active calibration mechanisms to provide, for example, long term statistics.
In another embodiment according to the present invention, active calibrator 208 resides in data director 220, or alternatively, is integrated into controller 205. There are several proprietary implementations of commercially available routers suitable to practice the present invention. One example of suitable active probes is the RMON probe. Cisco systems uses Service Assurance Agent (“SAA”) that is derived from the remote monitoring (“RMON”) probes to send out active probes. SAA allows routers to measure and report network-originated application round trip times. Although not every probe mentioned below is available in SAA for network calibration, one skilled in the art would appreciate how each of the following might be implemented to practice one or more embodiments of the present invention.
An exemplary active calibrator 208 can use ICMP (Internet Control Message Protocol) echo request or other ping-type probes, lightweight TCP-based probes, Sting probes, “pathchar” probes, lightweight probes using User Datagram Protocol (“UDP”) packets with a predefined TTL (time to live), traceroute probes, or other active probes that are suitable for use by active calibrator 208 in accordance with the present invention.
These probes that are received back by active calibrator 208 of
Exemplary passive calibrator 203 of
In another embodiment of the present invention, passive calibrator 203 examines the traffic stream in both directions (i.e., ingress and egress) and classifies each of the traffic streams into flows. Traffic flows are monitored within passive calibrator 203 according to the underlying protocol state (e.g., such as regarding TCP sessions) over time. For example, passive calibrator 203 classifies the traffic flow according to round trip latency, percentage of packets lost, and jitter for each of the traffic routes or flows. Such traffic route information is used to characterize the “end-to-end” performance of the paths carrying the traffic flows, which includes flow rates, and is aggregated into a series of network prefixes.
As described above, passive calibrator 203 is coupled to store, fetch and update traffic and route information stored in traffic repository 221 (connection not shown). Exemplary traffic repository 221 is a database configured to store and to maintain data representing traffic and route information that is useful to the end user employing a flow control system, such as system 200 of
Such feedback will comprise, for example, of types of traffic being sent, source addresses, destination addresses, applications, traffic sent by ToS or DSCP (“DiffServ Code Point”) setting (which might be integrated into a differentiated billing system), and volume of traffic. These statistics are fed into traffic repository 221 where, for example, a reporting engine or some other analysis process has access to them. The information stored in traffic repository 221 is data representing such traffic route characteristics arranged in any suitable data structure as would be appreciated by one skilled in the art.
In one embodiment, passive flow analyzer 330 performs passive analysis on the traffic to monitor current traffic flow characteristics so the controller can determine whether the monitored current traffic flow meets associated policy requirements. Export flow analyzer 331 performs passive analysis on exported flow records from a network device, such as from those devices (e.g., router) that advertise traffic type, source and destination addresses, and other information related to the traffic that it travels across service provider links. An example of such a network device is Cisco's Netflow™ product. In another embodiment, passive flow analyzer 330 operates in accordance to the passive flow analyzer described in the above-mentioned U.S. patent application Ser. No. 09/833,219.
Content Flow Analyzer 332 performs passive analysis of specific elements of data content, such as web site content. Export flow analyzer 331 and content flow analyzer 332 determine a set of relevant prefixes or a prefix list 334 that is associated with a specific user's policy. Prefix list 334 is sent as data representing such prefixes to an active detection process in the controller. Prefix list 334 can be one or more lists or data structures configured to store data representing performance and usage characteristics and are designed to receive a query, for example, by the controller. Once queried, the passive flow analyzer provides the one or more prefix lists, or portions thereof, to the controller for use in determining a policy violation, for determining which routes or path comply with the flow policy, which path is the optimum path for routing data, and the like. An exemplary prefix list can be generated by export flow analyzer 331 and content flow analyzer 332, as well as passive flow analyzer 330.
List 1970 includes aggregate flow information for each address 1920 or prefix. For example, record 1975 includes the following data: for address 220.127.116.11, this address was monitored four times during the sampling time interval (Δt)t with a total flow volume of 360 bytes. With record 1990 having a new prefix flag set (i.e., first time this address has been monitored), new prefix list 1980 includes address 18.104.22.168 having one occurrence (first time) over (Δt) interval. One having ordinary skill in the art should appreciate that other relevant data may be monitored and can be stored in list 1900. Moreover, the data representing address, occurrence, number of bytes, time interval, etc., can be used to manipulate the data such in a way that the controller can easily obtain it.
For example, the data stored within a list 1920 can be aggregated or grouped according to address or prefix. As shown in
Export flow analyzer 331 and content flow analyzer 332 also are configured to notify controller 205 when a previously unseen prefix has been added to the prefix list 334. New prefix notification signal 335 enables the control element 1005 to establish a new baseline performance for this prefix and to seed the routing table with a non-default route, or alternative route (i.e., non-BGP), if necessary. In one embodiment, export flow analyzer 331 and content flow analyzer 332 provide for monitoring of performance characteristics.
Content flow analyzer 332 is typically used when the main source of traffic flow 340 is web site or other content. Content source 341 can be configured such that special or premium content 342 that must be optimized can be identified by the flow control system by using, for example, an embedded URL 343. URL 343 redirects the client to a small content server running on the content flow analyzer 332. Content flow analyzer 332 receives a request for the small content element, which is generally a small image file (e.g., a1×1 GIF) and is invisible or imperceptible in relation to the main original content, and responds to the client with the small content element 344. Content flow analyzer 332 then stores or logs this transaction, and by using these logs, content flow analyzer 332 is able to perform aggregation and assemble content prefix list 334. The list 334 is passed along to controller 205, for example, for active service level monitoring and policy enforcement.
The small redirect URL acts to generate an HTTP request 420 in response to the small element of content. Content flow analyzer 432 sees this request 420 and responds 422 to it with, for example, a lightweight HTTP server 453. This server is fast and lightweight, and does nothing other than respond with the image file. The lightweight web server 453 logs the IP address of the client requesting the web page, and sends the one or more addresses to aggregator 454. Aggregator 454 aggregates, or collates, individual IP elements 424 into prefixes of varying granularity (e.g., /8 through /32) and also aggregates the frequency with which each prefix is seen over an interval of time.
That is, aggregator 454 classifies prefixes according to its frequency of occurrence and provides aggregated (i.e., grouped) prefixes 426 to prefix list generator 455. Prefix list generator 455 creates destination prefix list 428 according, for example, to a prefix's importance in relation to the overall operation of the system as defined by the aggregated or grouped prefixes 426. For example, each monitored traffic flow is examined to determine the performance characteristics associated with a destination prefix or address.
Aggregate prefixes 426 are generally classified in terms of flow frequency, and average or total flow volume. Prefix list generator 455 sends an updated to current prefix list 428 to controller 205 of
Referring back to
Destination prefix list 528 is generated by prefix list generator 552 by, for example, ranking and organizing traffic flow characteristics related to prefixes in order of relative importance. List 528 contains data representing an aggregation of prefixes and is organized relative to other prefixes, as determined by the system or an entity to ensure policy enforcement. For example, one or more prefixes can be ordered in terms of flow frequency and average or total flow volume in relation to other prefixes available to the overall system. Prefix list generator 552 sends updates to the current prefix list to controller 205 of
Passive Flow Analyzer 630 uses a copy of the traffic 602 via a passive network tap or spanned switch port, as shown in
Additionally, passive flow analyzer 630 is designed to be “application-aware” according to how each of the particular traffic flows is classified. Traffic can be classified according to the classifier described in the above-mentioned U.S. patent application Ser. No. 09/833,219. That is, passive flow analyzer 630 can inspect the payload of each packet of traffic 602 to interpret the performance and operation of specific network applications, such as capture and interpretation of the Realtime Transport Control Protocol (“RTCP”) for voice over IP (“VoiP”), for example.
Parser 651 is coupled to receive captured raw packets and operates to deconstruct the packets and retrieve specific information about the packet from each in the traffic flow. Exemplary parser 651 extracts information from the IP and TCP headers. Such extracted information from the IP headers includes source and destination IP addresses, DSCP information encoded in the ToS (i.e., “type of service”) bits, and the like. DSCP carries information about IP packet QoS requirements. Each DSCP defines the Per Hop Behavior of a traffic class. DiffServ has 64 code points so that it can define 64 different types of traffic classifications. TCP header information includes source and destination port numbers, sequence number, ACK number, the TCP flags (SYN, ACK, FIN etc.), the window size, and the like.
TCP elements parsed from the TCP headers are especially useful in determining whether a policy is being enforced, in terms of performance. An increasing amount of traffic, however, does not rely on TCP and instead uses UDP. UDP does not contain the necessary information to determine service levels according to conventional approaches.
To determine service levels to these destinations, the present invention might employ a statistically relevant amount of collateral TCP traffic going to the same prefix or a series of active probes to the same destinations, or have the analyzer parse deeper into the packet and understand the traffic at the application layer (e.g., layer 7). There are some protocols running on UDP that have very specific requirements that are different from most other data traffic on the network. These protocols are loosely classified as “real-time” protocols and include things like streaming media and Voice over IP (“H.323”). Packet loss and latency, below a certain level, are secondary concerns for real-time protocols.
Most importantly, however, is reducing the variance in inter-packet arrival times (i.e., network jitter). Many real time protocols such as H.323 report the observed jitter in back channel communication known as the RTCP (“Real-Time Transport Control Protocol”), which is used to distribute time-dependent media data via IP multicast with feedback. If passive flow analyzer 630 of
Correlator 652 operates to interpret and to group the packet elements (e.g., TCP and IP) from the packets to determine the current service level of the flow and then groups the packets into a specific traffic flow. Flows are reconstructed, or grouped, by matching source and destination IP addresses and port numbers, similar to the process of stateful monitoring of firewalls. Correlator 252 determines the current service level by measuring several traffic characteristics during a TCP transaction. For example, correlator 252 determines the round trip time (“RTT”) incurred on a network, and hence, this serves as a measure of latency for the network traffic.
Correlator 652 of
Packet loss is calculated, as a percentage, by correlator 652 by maintaining the state of all of the retransmitted packets that occur. From this value, correlator 652 calculates percentage packet loss from a total count of segments sent.
Correlator 652 also determines SRT 792 of
Correlator 652 also determines SRT 892 of
Referring back to
Referring back to
The granularity of the destinations is the same as the granularity of changes that can be made in the routing table. Nominally, flow control system of
Aggregation techniques employed by aggregator 653 are used to maintain the system 200 of
Usage collector 1215 also operates to reconstruct provider billing records. Usage collector 1215 accepts provider configuration information 1271 related to each NSP connection. This NSP configuration information 1271 details provider interfaces on the various routers 1272 (e.g., egress routers), provider next-hop IP addresses traceroute probes (to verify the current provider in use with trace probes), billing period start and end dates, circuit bandwidth for calculating the utilization and price per megabit/sec, minimum bandwidth commitment, burstable rates, provider sampling interval, provider billing algorithm, a utilization alarm threshold and the like.
In operation, exemplary raw collector 1274 sends a query 1290 (e.g., SNMP) to collect interface raw byte counters from routers 1272 on each of the provider circuits at a specified sampling interval. Provider circuits include paths, pipes (virtual or physical) T1 and the like. Raw collector 1274 places the raw byte counters 1280 into persistent storage for later reporting and analysis. Raw collector 1274 sends the raw information to two other components: utilization monitor 1275 and bill reconstructor 1276.
Utilization monitor 1275 calculates the ingress and egress circuit utilization for each provider using the raw byte counts and the NSP configuration information 1271. In one example, NSP configuration information 1271 includes the bandwidth of the provider's circuits. Utilization information 264 includes data representing utilization trends for use with short range forecasting models (e.g., ARIMA, exponential smoothing, etc.) such that utilization monitor 1275 can determine whether bandwidth is trending up or down (i.e., increasing or decreasing in size) for a given service provider.
Bill reconstructor 1276 uses the billing information from NSP configuration data 1271 to reconstruct the current provider billable rate for the current billing period. Billing information includes information explaining the methods that specific providers use to calculate costs, such as a billing rate. Such methods of calculating bills for using a network provider are well known in the art. Bill reconstructor 1276 applies similar provider billing methods to the raw byte counters from raw collector 1274 to generate the bill and related billing rates, etc. The generated bills, which are mapped into dollar amounts, are typically estimates since the sample times between the provider and usage collector 1215 will not match exactly. Bill reconstructor 1276 will send billing information 1261 to controller 1202 for use in peak avoidance and least cost routing. Peak avoidance is defined as a method of avoiding using a path or path segment at a higher a billing rate, such as shown in
Additionally the information can be sent to controller 1202 for use in the least cost fix method of selecting the cheapest if performance is of no consequence. That is, controller 1202 uses data from billing message 1261, including billing rates, to determine an alternate route based in part on a route's free bandwidth (i.e., route does not incur additional cost to use), in accordance with the flow policy.
Referring back to
Once data director 220 receives this information, direct or 220 informs controller 205 that route change has been implemented. Thereafter, controller 205 communicates signal 230 back to passive calibrator 202 to clear its state and to resume monitoring the destination. The destination is monitored to ensure that the updated route of the routing table, or path, meets minimum service levels (e.g., no violations of SLA, or no unacceptable deviations from agreed upon performance metrics as defined by the associated flow policy).
In one aspect, configuration element 211 resides in a route server. In another aspect, configuration element 211 resides in a router and is configured to modify a route map or table. In yet another aspect, configuration element 211 is adapted to provide configuration information, or a routing table. In still yet another aspect, the route information is stored within the configuration element 211 according to whether it is related to inbound or outbound traffic.
In one embodiment, the routing tables are received into route server 1391 by way of one or more default BGP4 feeds 1392 into BGP4 Engine 1382 from a full set or sub-set of the local transit providers. BGP4 Engine 1382 integrates, or merges, all of the routes into a single BGP4 routing table 1383 of the best available routes. In another embodiment, route server 1391 maintains an iBGP session with all of the internal BGP capable routers rather than maintaining the BGP4 sessions, as shown in
Configuration element 1384 is designed to receive one or more BGP4 routing tables 1383 from BGP4 engine 1382 and is adapted to receive one or more control signals and data resulting from the control processes of controller 1305. In operation, configuration element 1384 receives, from controller 1305, the necessary routing changes to be implemented in default routing table 1388. Then, configuration element 1384 incorporates one or more changes in modified routing table 1389.
Thus, configuration element 1384 operates to modify BGP4 routing table 1383 and to generate one or more modified BGP4 routing tables 1388. Modified BGP4 routing table 1388 includes changed routing 1389, advertisements of more specific routes, etc. New modified BGP4 routing table 1388 is then fed to all BGP clients in the network, which then is used to guide traffic to the destination.
Controller 205 of
The active calibrator and the passive calibrator provide performance characteristics. Regarding the active calibrator, controller 205 initiates active calibration by request active probing. The active calibrator sends one or more calibration probes on probe path 207 out into the one or more data networks. The returning probes on probe path 207 provide information back to controller 205, which contains the identities of available paths and performance information related thereto.
Regarding the passive calibrator, controller 205 is designed to receive real- or near-real time network performance characteristics (i.e., loss, latency, jitter, etc.) from passive calibrator 230 as a monitor in traffic flows in which it has access. After controller 205 provides a routing change, or update, to configuration element 211, it also communicates a signal 230 to passive calibrator 203 when an updated route change is made to a specific destination. Signal 230 initiates the clearing of the state of passive calibrator 203 so that the calibrator resumes monitoring the specific destination to ensure that the updated route of the routing table, or path, is flow policy compliant. Clear state signal 338 of
In one example, controller 205 operates to interpret the aggregated flow data over an interval of time for each of the groupings of destination prefixes. And if a policy violation occurs, controller 205 determines which of the alternate routes, or paths, are best suited for the prefix or traffic type associated with the current traffic flow. Controller 205 then sends the necessary routing changes to configuration element 211. That is, controller 205 resolve policy violations relating to non-compliant network performance characteristics, in accordance with the associated flow policy. This process is repeated until the policy violation is resolved.
In another example, controller 1202 of
To resolve usage-type policy violations, controller 1202 is configured to receive routing tables, for example, to determine which of the current traffic flows or routing of data on certain paths, or path segments thereof, are congested (i.e., loaded) with respect to a particular provider path or paths. Controller 1202 also is designed to receive data representing flow volumes for each of the alternate provider paths to determine which sub-set of flows of a set of traffic flows to or from a given destination prefix are in compliance with the associated flow policy in terms of traffic flow volume.
An exemplary controller of the present thus is designed to obtain information related to the performance and usage of data networks and effectuate corrective action to effectively and efficiently route data over paths or segments of paths that meet at least associated policy requirements.
The following discussion relates to flow policies and the application of such policies in resolving policy violations and in enforcing the policy requirements or metrics. Referring back to
Detection is defined as the techniques or mechanisms by which flow control system 200 determines which traffic should be acted upon in response to a policy violation. The traffic flow can be identified, by name, by source or destination addresses, by source or destination ports, or any other known identification techniques. For example, a policy can be associated with only one prefix. That is, system 200 will monitor the traffic flow to and from a specific prefix, and if necessary, will enforce the associated flow policy in accordance with its requirements. Further regarding detection, a policy defined for more specific prefixes can take precedence over more general prefixes. For example, a policy defined for a /24 will take precedence over a /16 even if the /16 contains the specific /24.
Performance is a policy requirement that describes one or more target performance levels (i.e., network/QoS policy parameters) or thresholds applied to a given prefix or prefix list. Although more than one performance-based policy requirement may be defined, in this example only a single policy is applied to a given prefix or prefix list. Exemplary performance requirements include loss, latency and jitter.
Moreover, such requirements can be configured either as, for example, an absolute, fixed value or as an exponentially weighted moving average (EWMA). An absolute value establishes a numerical threshold, expressed as a percentage or in time units over a configurable time window. The EWMA method establishes a moving threshold based on historic sampling that places an exponential weighting on the most recent samples, thereby asserting a threshold that can take into account current network conditions as they relate to historic conditions.
Additionally, process control and statistical quality control techniques can be used to determine when the underlying network performance has changed. These techniques, used effectively in manufacturing, can determine what the upper control limit should be for a given destination at any time, and then notify the system when the ‘quality’ of the network performance has changed or falls outside these control limits. For example, several absolute thresholds can be used to effectively and simply overcome the non-linearity of the system. Network performance generally falls into one of several operating floors defined by the proximity of the destination (e.g., metro, regional, continental, inter-continental). These operating floors can be defined by the running mean RTT. Each operating region has its own upper control limit or threshold. When the mean RTT falls above the UCL (upper control limit) for several data points, a violation is triggered.
Cost is expressed in the policy definition in terms of precedence and whether the policy is predictive or reactive. Costs are characterized by usage collector 214 of
Typically, both cost predictive and reactive requirements result in a binary decision (i.e., a circuit or path, for example, is either in compliance with or in violation of a flow policy). In the case of predictive cost, the transit circuit is either in compliance, or soon to be violation of a flow policy. Regardless, an action must be taken to resolve the situation, unless cost is preceded by performance (i.e., performance requirements are to be addressed prior to making a cost-based change).
Precedence is a policy requirement that describes one or more target usage or utilization characteristics or levels. Precedence includes provider preference and maximum utilization (i.e., load) requirements. The provider preference requirement is, for example, an arbitrary ranking of providers that is used when an action must be taken, but when two or more transits may be selected in order to enforce the policy. The flow control system can automatically set the provider or path preference requirement if it is not configured explicitly by the system's operator. This requirement is then applied as a tiebreaker in deadlocked situations such that the provider with the highest preference wins the tie and thus receives the diverted traffic flow.
The maximum usage requirement can be used as either may also be used as an actual operational threshold not to be exceeded or as a tiebreaker. Maximum usage is configured, for example, in the transit provider section of the configuration and takes either a percentage argument (i.e., in terms of available bandwidth), or alternatively, can be set as an absolute value in terms of Mb/s (i.e., not to exceed available bandwidth).
The following is an example of a policy used with a controller to determine whether the specific policy is in compliance, and if not, to determine the course of action.
For example, consider the following policy is used for a particular traffic flow:
Suppose that traffic flow associated with prefix 22.214.171.124/24 is currently carrying traffic at 240 kbits/sec, and is reached via provider 1 of 3. Provider 1 is currently carrying 2 Mbits/sec and has a minimum commit of 5 Mbits/sec.
The controller of the flow control system using the policy can monitor the alternate traffic routes, or paths, and can determine the following flow characteristics as they relate to the providers:
In this case, ISP1 is in a violation state since loss of 5% exceeds the maximum loss requirement of 2% and since loss has been designated with the precedence of 10, with 50 being the lowest. Corrective action must be taken. The policy will be enforced without latency or loss information (i.e., because there is, for example, no visibility into the performance of the other links). In this case, the controller may initiate active probing using the active calibrator to determine whether the other ISPs (including ISP2 and ISP3) are in compliance. Alternatively, the controller might determine the course of action based on the next parameter in the policy where the requirement is known (e.g., cost in this case). Since ISP 2 is in compliance and ISP 3 is not, ISP 2 would be chosen by the controller. If the two were both in compliance, the controller would go to the next ranked requirement, which is MaxUti1. If this is the case, ISP2 is still selected.
In summary, the policy, such as the above exemplary policy, is input into the controller 205 of
In order for controller 205 to handle peering connections, controller 205 communicates with the data director 220 to retrieve reachability information (i.e., routing tables) for the specific prefix that is about to be changed. In the case of transit circuits, controller 205 uses active calibrator 207 to determine reachability information (i.e., routing tables) for a given destination by, for example, sending active probes to the destination and then waiting for the response. Although peering connections are often unreachable, it is possible for active probes to succeed since some providers may not effectively filter traffic at a peering point and instead rely on an honor-like system to ensure that only traffic to those advertised destinations is received.
Therefore, in the case of peering, controller 205 must look in the routing table for an advertisement of that destination before moving traffic to a peering connection. Referring to
SPRE 1800 includes flow inspector 1812 and a dynamic information base (“DIB”) 1813, where flow inspector 1812 and DIB 1813 are adapted to communicate data with each other, uni- or bi-directionally. DIB 1813 includes one or more information bases (e.g., data tables), such as a routing information base (“RIB”) 1820, a forwarding information base (“FIB”) 1822, a network address translation information base (“NIB”) 1824, or other suitable information bases or data structures used to maintain routes for active flows while implementing a route change. Such information bases are deployed in any storage device or memory known in the art, such as a DRAM, a floppy disk, or the like. Additionally, information bases use any data structure for associating data representing routes to network elements receiving traffic and data flows related to such routes, for example, and like routing or forwarding information used to effectuate routing changes.
DIB 1813 and its constituent information bases are designed to communicate uni- or bi-directionally with controller 1805 and data director 1814. Data director 1814 operates to forward the traffic flow (i.e., one or more data packets associated with a remote data address) originating from the local source address to a data network associated with a path or route that meets or exceeds a specific flow policy. Prior to making a route change, controller 1805 performs a look-up using data director 1814 by initiating a look-up with signal 1834. For example, the controller looks up the current route or path that the data director 1814 to route egress and/or ingress traffic. Such route information is typically stored in one or more routing tables and is communicated to controller 1805 as provider look up data 1832. A provider is a data network service provider and has an association stored in the routing table to route data by such associations.
As described herein, a path (or route) generally is an egress path from a first point, such a source address, to a second point, such as a destination address, or an ingress path from the second point to the first point. A path therefore is modified by rerouting the data via an egress portion of the path from the first point to the second point, or alternatively, by rerouting data via an ingress portion of the path to the second point from the first point. One having ordinary skill in the art should appreciate that a portion of a path includes one or more constituent segments of either an egress or an ingress path between the first and second points (i.e., a portion includes every segment of an egress or an ingress path). Also, although a local source address can be associated with a first point and a remote destination address can be associated with a second point, it should be understood that defining a first point as either a source or destination is a matter of convention.
Data director 1814 and its components are discrete network elements or devices and include routing engine 1826, forwarding engine 1826, NAT engine 1826, or the like. Each engine can be associated with at least one information base, and is adapted to forward one or more data flows out into the data networks of the providers and to receive traffic from a remote destination address, regardless of whether the data flows are on the same or different ingress path as the egress traffic, for example.
In operation, SPRE 1800 generally alters the data flow outbound from the local source network where it originates and/or inbound from the remote destination address into the local source network. Exemplary SPRE 1800 employs network translation (“NAT”), virtual local area network (VLAN), and other suitable data forwarding and networking techniques, some of which is known in the art. Moreover, SPRE 1800 can use any routing protocol suitable for modifying the routing of local data traffic flow, for example, out from a local source network and into one or more network service providers.
NAT is generally used to associate a second set of addresses for public traffic with private traffic. More specifically, when a host on the “private” network sends a packet (e.g., IP packet) to a data network (e.g., the Internet) via a network device using NAT (i.e., NAT-enabled), the network device keeps track of the actual “private” address of that host, but substitutes an assigned “public” address into the packet before it is sent into the data network. When a reply comes back from the data network, the NAT-enabled device, such as a router, restores the actual address before sending the reply to the host. In some firewall applications, NAT is used to determine whether the inbound traffic from the data network should be permitted to enter the local source network.
According to a specific embodiment of the present invention, NAT is employed to modify or to translate the source address of the data traffic (i.e., of each packet) flowing out from a local source network. With a translated source address, that the inbound or ingress path returning traffic back to the local source network is different that the egress path. Moreover, the current ingress path or a portion thereof, is changed to a next ingress path or a portion thereof, over time or after a route change. For example, the source address of an IP datagram is modified to return through a different NSP than the NSP associated with the outbound traffic. An NSP is referred to as a data network herein.
VLAN technology is typically used to associate physical network elements (e.g., routers, switches, etc.) or computing devices (e.g., a network of computers) with virtual equivalents through software rather than hardware. A corresponding tag identifies each unique network element, such as a border router coupled to a first level data network. A suitable tag associating a network element with a data flow is an IEEE 802.1q VLAN tag and is well known in the art. Such VLAN tags are used to route data traffic (e.g., packets) to at least one data network in a group of data networks. For example, a unique VLAN tag is used to forward data traffic from a source address, S, to either NSP 1 or NSP 2 to effectuate paths A or B, respectively, as shown in
SPRE 1800 can use any suitable routing protocol, such as BGP4, to modifying the routing of local data traffic flow as described above. Although SPRE 1800 may function within an environment in which eBGP or iBGP is used, BGP is not necessary to practice the present invention. One having ordinary skill in the art should appreciate that BGP4 is but only one way to influence the path associated with egress data traffic from the local source network to a destination (e.g., remote destination address) and other suitable routing techniques are in accord with the present invention.
Exemplary SPRE 1800 can be implemented similarly as shown in
Exemplary SPRE 1800 includes flow inspector 1812. Flow inspector 1812 operates to determine and to indicate which of the traffic flows (i.e., local data flows) from the local source address is active or inactive. Such traffic flows generally flow out from local source network via one or more data networks to a remote destination address, especially if a multi-homed scheme is employed. Flow inspector 1812 also functions to provide a sub-set of all available flows, or alternatively, a super-set of specific individual flows (i.e., where each individual flow is denoted by a specific destination address) to one or more information bases in DIB 1813, the sub-set including one or more active flows. Moreover, flow inspector 1812 determines when a particular traffic flow is inactive, and upon doing so (or thereafter), communicates a request to withdraw inactive flows in the one or more information bases in DIB 1813. An inactive flow is withdrawn by, for example, deleting one or more inactive data flows associated with one or more remote destination addresses, thus freeing up storage locations in the information bases, among other benefits.
In one embodiment, flow inspector 1812 of SPRE 1800 determines whether a flow is active or inactive by performing stateful inspection of one or more constituent elements of an IP datagram, such as individual bits, fields, or the like (e.g., portions of an IP header). Exemplary flow inspector 1812 inspects the TCP portion of each packet in a flow associated with one or more remote destination addresses. SPRE 1800 typically initiates the process of stateful inspection upon receiving a route change request from controller 1805. Stateful inspection of data flows, such as local traffic flows, include examining one or more constituent elements of a data packet over time and comparing the state of each data packet of a data flow associated with a destination address, where the data flow is compared against one or more inspection rules. Thereafter, flow inspector 1812 is configured to initiate action in response to the result of such comparison. For example, flow inspector 1812 identifies each flow as either active or inactive according to the inspection rules tested against. Inspection rules, for example, include one or more criteria or requirements used to determine whether a flow is to be deemed active or inactive.
For example, flow inspector 1812 operates to select a specific protocol, such as TCP, and to examine TCP session-related characteristics of each packet. Such TCP characteristics include a source address, a destination address, a source port and a destination port as well as protocol, time stamps, and the like. One having ordinary skill should appreciate that such examination is performed according to other protocols, such as UDP, etc. In another exemplary flow inspector 1812, the well-known port numbers of the Internet (e.g., RP94) are used to examine (i.e., demultiplex or fragment) the packets to determine which protocol is being used based on TCP port numbers. This information can be used to determine which end of a TCP connection is the server in a TCP session. Exemplary flow inspector 1812 is designed to perform flow reconstruction, similar to flow reconstruction performed by the correlator of the above-described passive analyzer. Yet another exemplary flow inspector 1812 performs flow reconstruction with respect to layer 4.
Over time, flow inspector 1812 reconstructs the flows and stores the states (i.e., TCP characteristics) of each flow according to a key, such as a specific remote destination address or super-set thereof. The states are stored in a state database, such as a hash table, for creating an index for one or more records in a database. An exemplary hash table stores state data representing the states or values of TCP characteristics for each flow in an appropriate data structure. Such a data structure is used to associate the state data with a specific remote destination address or address super-set (i.e., key), which is designed for quick data retrieval. Hash table data structures are well known and one having ordinary skill should appreciate how to implement hash tables in accordance with the present invention.
Flow inspector 1812 is designed to provide DIB 1813 with a number of active flows 1816 that have been defined as active for one or more remote destination addresses. DIB 1813 and each of the information bases includes logic for communicating with flow inspector 1812 so as to receive, for example, requests to withdraw inactive flows 1815 associated with a particular information base in DIB 1813. Moreover, DIB 1813 and each of the information bases, in one embodiment, include logic for communicating with one or more engines of data director 1814. In another embodiment, some of the information bases include logic for communicating with one or more engines of data director 1814. In particular, each information base is designed to provide at least one default or current path 1825 for each packet (or data flow) associated with a specific remote destination address, typically in response to a request for a look-up 1825.
Data director 1814 and/or its constituent engines are adapted to receive one or more packets of a data flow and operate to compare the associated remote destination address of the packets to data representing an interface (e.g., physical or virtual), translated address, or other network device identifiers used to govern an ingress or egress path. Network device identifiers are associations used to govern the flows of data over varying egress and ingress paths. For example, a specific virtual router facility is associated with a VLAN tag as well as a variable stored in a data table, such as VLAN 101 of
Returning back to
One having ordinary skill should appreciate that DIB 2100 contains as many records and fields (e.g., RIB 2102, etc.) necessary to practice the present invention. In another example of DIB 2100, each of the information databases in DIB 2100 is separated into separate data structures or network devices such that an exemplary RIB includes remote destination addresses and physical interfaces without the FIB, NIB, and other information. Similarly, separate NIBs and FIBs can be constructed to exclude the other network device identifiers. In another exemplary DIB 2100, rather than the flow inspector being configured to include logic to perform comparisons of each flow against a set of inspection rules, DIB 2100 is adapted to include logic and a set of inspection rules for determining whether a flow is active, and how best to effect a route change by modifying one or more information bases without disrupting the associated traffic flows.
In another example of SPRE 1800, each separate RIB, NIB, FIB, etc. is designed to provide specific network-identifier information (such as a physical interface, a virtual interface, etc., as identified by EG#, VLANxxx, NATxxx, etc.) to a specific engine residing within or without a network element. That is, routing engine 1826, forwarding engine 1828, and NAT engine 1830 is configured to communicate exclusively with RIB 1820, FIB 1822 and 1824, respectively, such as depicted in
Referring back to
For example, flow X of
Prior to a route change, for instance, flow X is actively flowing out through a physical network element identified as EG1. The data director or routing engine then provides this information to the controller or the RIB 2102. If flow X, however, is actively flowing out through a virtual network element identified as VLAN101, then data director or forwarding engine provides this information to the controller or the FIB 2104. Furthermore, if flow X is actively flowing into a local source network through a physical, a virtual network element or the like, the address of such an element is identified as NAT 10. This translated address then is provided to either the controller or NIB 2106.
Upon receiving a route change request from a controller, a particular information base or DIB 2100 requests a look-up of the current or default routing related to a super-set of remote destination addresses affected by the route change, which in this example is 0.0.0.0 /24. The respective engines, or alternatively the data director, then inserts 0.0.0.0 /24 into DIB 2100 as record 2150.
The flow inspector inserts the active routes 2120, such as 0.0.0.1 /32, 0.0.0.2 /32, 0.0.0.3 /32, etc., into DIB 2100 and associates those routes with the routing information (e.g., from a routing table) used by the data director 2130 to route the traffic flows. By doing so, the active flows will continue to flow according to the routes prior to a route change. As these individual active flows are deemed inactive, DIB 2100 withdraws, or deletes, the active flow record from the DIB.
In one embodiment, DIB 2100 operates to withdraw the original routes related to associated network element or translated address identifiers (e.g., EG1, VLAN101, NAT10, etc.) for one or more (i.e., a subset or super-set) of routes to be changed. In this example, routes associated with 0.0.0.0 /24 are withdrawn at 2132. Then the controller provides a route update 2140 associated with the route change at 2134. Further to this example, after the update, ingress and egress traffic described by 0.0.0.1 /24 is thereafter be routed out from a local source network according to EG2 or VLAN 102, or can be routed into the local source network according to NAT 20. Traffic flows will continue to be routed accordingly until the controller decides it is necessary to make another route change based upon performance and usage characteristics, as described herein. In another embodiment, a controller makes such routing changes to either the inbound or outbound traffic, or both, according to routing algorithms known in the art.
The following discussion relates to
In this example, flow inspector 2212 checks for and captures each active flow having a prefix, such as TCP layer 4 flows, that have not received a FIN or RST. These, and other similar flow types, are interpreted as being active in the TCP state machine. Thus, the source address of a flow cannot be changed without disrupting the flow. Each of the active sessions is associated with a super-set of addresses, or prefixes, and is recorded. Thereafter, flow specific routes, such as /32 entries, are inserted into RIB 2220. Flow inspector 2212 includes a set of rules for preserving active flows by, for example, storing them in a hash table before making a pending route change.
After the route change has been made, flow inspector 2212 monitors each individual active flow (i.e., TCP session) to detect whether it has terminated (i.e., become inactive). For each inactive flow, flow detector 2212 is configured to withdraw the specific routes (e.g., /32) from RIB 2220. Thereafter, flow detector 2212 continues to monitor the states until all active sessions on the previous route have ended and all specific flow routes (e.g., /32) have been withdrawn. RIB 2220 then places the update route change into a routing table at the routing engine.
Routing engine 2226 operates to maintain one or more routing tables and examines RIB 2220 to determine how best to forward the associated traffic flow according to the routing information in the information base. That is, routing engine 2226 uses RIB 2220 to forward the data traffic flows over one of a number of physical interfaces toward an associated network element (e.g., egress gateway), such as routers. An exemplary routing engine 2226 operates as a layer 2 forwarding mechanism and is configured to perform at least one route lookup into RIB 2220, where look-ups are performed to identify the current path or paths or portion thereof of an active flow.
In another exemplary SPRE 2200, a default route associated with a least-cost network service provider, for example, is installed in the routing tables as well as in the RIB 2220. SPRE 2220 is coupled to at least one of the available physical egress interfaces from the local source network, such that data traffic flows, by default through the least-cost NSP until the controller determines whether a route change is required. The default route keeps SPRE 2200 from responding and making routing decisions based upon BGP advertisements in to which it may be exposed.
In yet another exemplary SPRE 2200, SPRE 2200 is configured to communicate with its environs based on any available protocol, such as iGP (e.g., Open Shortest Path First protocol) or eGP (e.g., BGP), and is further configures to update RIB 2220 based on route advertisements that are received from those protocols as well as any other protocol known in the art.
Forwarding engine 2326 is designed to tag the outgoing data flows with specific identifier information, such as layer 2 information, and more specifically, VLAN tags. Forwarding engine 2326 examines FIB 2320 for information to determine which VRF is to be used to send the traffic out on. VLANs, according to one embodiment, logically are associated with one or more physical interfaces, where an exemplary SPRE 2300 is coupled to any one of a number of virtual routers (e.g., VRFs), where each virtual router is logically associated with a physical interface.
Downstream routers are implemented to route data traffic 2450 forward from SPRE 2400 onto available network service providers according to BGP, for example, or any other routing protocol. In this example, traffic flows enter the downstream routers according to a specific NAT address, as translated by NAT engine 2426, and according to an ingress route change request generated by a controller.
In another example, SPRE 2412 operates with a flow inspector 2412 that is similar in function and nature to flow inspectors 2212 and 2312 of
Structurally and functionally,
Exemplary NIB 2420 includes one or more NAT addresses to be applied in a similar fashion as described herein in connection with a routing information base (i.e., RIB). Flow inspector 2412 also monitors all active flows for a specific prefix change stored in a network address translation information base, or NIB 2420. NAT engine 2426 inserts temporary, but specific route entries, to maintain a current flow policy while one or more flows are still active.
A typical application for NAT is where the originating source address in a private IP address space is translated into an address for use in public routable space. In this instance, NAT is used to significantly change the way that traffic is sent toward the originating network. NAT is acceptable for most applications and the policy that enforces NAT to prevent NAT from breaking the data packets of the relevant applications for which NAT is unacceptable. NAT, however, maybe overridden, and thus not useable, in some applications on the Internet. Such applications include Voice over IP (i.e., H.323 protocol), and security protocols, such as IPSec (IP Security Protocol) used to protect client information over IP. These applications might be denied the ingress path diversity provided and thus be forced onto a sub-optimal network path. Hence, according to yet another embodiment of the present invention, a specific router or network element is reserved to serve route-sensitive data packets and their attendant applications. For example, a single router may be reserved, using NAT, for example, to receive ingress data traffic into a local source network, if it is crucial that the data packets not be routed according to route control techniques described herein.
Additionally, changing the source address in the middle of an active flow might disrupt that flow. Other embodiments of the present invention described herein are implemented to overcome the NAT's disruption of active flows. For example, an exemplary controller might be implemented to verify that no active flows to that destination are in progress before the controller initiates a route change where NAT is applied to a traffic flow to affect its return path back into the source network.
Further to the specific embodiment of SPRE 2500, an associated traffic flow packet flows to router R1 2510 via interconnection 2512. At R1, a NAT engine distributed within the router functions to modify the source address of the data packet such that the ingress traffic flow back into the local source network may enter through any other NSP, such as NSPn, instead of through NSP1. A similar NAT engine is disposed in Rn to provide the same routing control of the ingress path if a data flow enters the Internet, for example, via NSP2.
Such a data packet then flows to router 2606. Router R0 2606 is configured to tag the NAT-altered data packet 2604 with a VLAN tag 2605 such that the data packet will enter NSP1 or NSP2 (or another NSP) according to the information within the VLAN tag. For example, consider VLAN tag 2605 is associated with a virtual interface linked to NSP1. That is, VLAN 2605 is designed to forward data flows out from R0 2605 via VRF1 2608. VRFn 2609 need not be used if VLAN tag 2605 is associated with VRFn 2609 of R0 2606. The packet 2604 flows out of the local source network via R0 and into NSP1. The packet, however, may return on an ingress path different from the egress path. Control over both egress and ingress routes for each data flow are but one aspect of the present invention, as embodied within an exemplary SPRE.
SPRE 2731 is configured to accept ingress route change information 2718 from the controller 2705 and uses a NAT engine 2717 to modify one or more NAT addresses 2720 on a per prefix basis. NAT engine 2717 reports on the prefix per provider information from the NIB with a prefix list by ingress NSP message 2732, for example, when requested by controller 2705
Additionally, SPRE 2731 accepts egress route change information 2791 from controller 2705 and uses route server 2788 operating with BGP, as described in
Although the present invention has been discussed with respect to specific embodiments, one of ordinary skill in the art will realize that these embodiments are merely illustrative, and not restrictive, of the invention. For example, although the above description describes the network communication data as Internet traffic, it should be understood that the present invention relates to networks in general and need not be restricted to Internet data. The scope of the invention is to be determined solely by the appended claims.
In the foregoing specification, the invention is described with reference to specific embodiments thereof, but those skilled in the art will recognize that while the invention is not limited thereto. For example, other network topologies and protocols, such as ATM, UDP, PVC, or any other communications protocol where information is transported in one or more packets, can be utilized in conjunction with some embodiments according to the present invention.
Various features and aspects of the above-described invention may be used individually or jointly. Further, although the invention has been described in the context of its implementation in a particular environment and for particular applications, its usefulness is not limited thereto and it can be utilized in any number of environments and applications without departing from the broader spirit and scope thereof. The specification and drawings are, accordingly, to be regarded as illustrative rather than restrictive.
Citat från patent
Hänvisningar finns i följande patent